DxO With Older Image Files

John Lehet at his most recent solo photo show in
New Hampshire, October 2015

-

Written by John Lehet

When you and I last spoke in person, Mike, at my last show in New Hampshire, you expressed surprise that I used DxO. I use it mostly for the reasons you point out in Tuesday's post, "Ah, the Night Sky (Old FF Dynamic Range)." While sensors have certainly improved a good bit, demosaicing and noise reduction algorithms have improved more quickly. So if you shot RAW in the past, your images can have an improved second life. Some of the images of mine that you saw hanging were made with older, lesser sensors and lenses, but I was able to coax some quality out of them.

I know that the other premium raw processors have their adherents, and probably for a reason, but I  settled on DxO as soon as I tried it. It was the first time I could reduce noise without degrading image quality. Also, DxO's deconvolution-based approach to lens correction corrects for lens flaws in lesser lenses of my relative youth.

Not only are my older RAW images from 2004–20xx getting essentially better sensors; it is also as if they are getting re-exposed with better lenses. I have to admit that I was ignorant in settling for some pretty mediocre Nikon zoom glass early on, especially on some trips where I made some good images. DxO doesn't help the bad bokeh, but it improves the sharpness and microcontrast as well as the noise so much that I have taken the time to re-process all the images on my site (over 250 of them!) since I started using DxO about a year and a half ago. This has taken a ridiculous amount of time, but it has made some much better images.

When I go to rework an image, I'm often quite startled at how much better the modern version is, to the extent that I'm embarrassed to have my old ones still on the site. They are mostly gone now. The benefit of using DxO is far less dramatic on the files from the D800e with primes than on 2007 files from the D200 with a zoom. The DxO difference does show up in the Micro 4/3 files even with good modern lenses, and it's possible those files will get even better in the future as software improves.

Robert Harshman said in the Comments yesterday that DxO, PhotoNinja, and ACR can each "do things the others cannot, and based on needs one may be much better than the other." I'll second Robert's comment: try DxO on older images.

John

©2015 by John Lehet, all rights reserved

Links in this post may be to our affiliates; sales through affiliate links may benefit this site.

TOP's links!

(To see all the comments, click on the "Comments" link below.)
Featured Comments from:

No featured comments yet—please check back soon!

Comments

[Howard]

I also am impressed with how DXO dramatically can improve microcontrast, sharpness, and noise compared to lightroom when processing raws. The degree of improvements vary from image to image and lens to lens, though generally the DXO files are substantially better. I have been using DXO Optics Pro for many years and now use it to process almost every image I want to print before printing them. I now use primarily the m4/3 format, though I've used many formats over the years, from large format film (Velvia) to full frame digital (the Canon 5D, which files DXO also improved substantially over lightroom) to m4/3.

[Floyd Takeuchi]

An enthusiastic second to John Lehet's recommendation of DxO. It gives me predictable high quality results, particularly for printing for exhibition. I use DxO's default settings.

I usually print 16x20 and 30x36 for exhibition prints, and have even printed one file 98"x56" on canvas for a hotel commission, all from Nikon D700 files. I use an experienced printer, who I trust to tweak the files if necessary. But he usually does not have to -- and my post-processing skills are pretty basic.

Like John, I've also had very good results with Micro Four-Third files, printing them 16x20 for gallery sale. And my printer says I could easily go bigger, a testament to both the Olympus M43 sensor, Olympus and Panasonic prime lenses and DxO.

I've found the DxO staff to be very responsive to questions, particularly inquiries about camera-lens modules, which John refers to. My only "complaint": I wish DxO was a bit faster getting some modules to users. The positive: it keeps me from rushing to add new cameras/lenses to the bag.

[Manuel]

I concur. I have chosen DxO over Lightroom after one month of intensive trials and side-by-side (well, almost) comparisons. From then on it excelled at correcting lens distortion, geometry, and noise from Raw files, but it can do wonders with JPEGs and even with scans of old pictures.
And the good news is that, if it recognizes the camera and lens, you might be satisfied with DxO's own presets. I found out it saved me a lot of fiddling with sliders, which Lightroom didn't. As John asserts, it can turn a middling image taken with a mediocre lens/body combo into a high quality photograph. Having processed many pictures taken with the Olympus E-P1 and the rather mediocre 17mm-f/2.8 and the appalling 40-150-f/4-f/5.6 zoom, DxO never failed to correct most of their flaws. Little short of magic.
There are some minor caveats. If you overdo it, or leave the programme to its own devices, chances are you'll end up with HDR-esque images, which is not to everyone's taste. It doesn't convert images to grayscale, which I find to be the only acceptable way of converting Raw files to black and white. And it won't even open grayscale TIFF scans from black and white film. (Although, strangely enough, it not only opens but also makes a terrific job out of colour film TIFF scans: go figure!) But the latter two are personal issues; if you work strictly with digital cameras and don't mind the relatively slower workflow, I wholeheartedly recommend DxO.

[Steve Rosenblum]

I have used and liked DXO Optics Pro for awhile--it can at times do remarkable things that other programs cannot. One drawback is that it's camera/sensor specific algorithms will not work with most Fujifilm X-tran files--the program only works with files generated by traditional Bayer matrix sensors. As I switch from my m4/3 gear to Fuji, DXO Optics Pro is a tool I will miss.

[Paul]

Hi Mike,
I'd love to hear Ctein's view on this post.
Wishing you a Merry Christmas.
Paul.

[Joe Holmes]

For many years, I ran every single image I processed first through DxO Tools, because it's brilliant at fixing lens distortion and chromatic aberration.

But when one of Nikon's latest cameras was released a couple years ago or so (D800? D810?), DxO was late updating the application, so I was forced to switch to Nikon's NX-D for a month to perform the same fixes (note: the interface is awful). Then, since I was used to the NX-D routine, once DxO was updated I decided to do some tests to see how the two tools compared. I ran quite a few images through both tools and compared at 100%.

I discovered that DxO's RAW processing was inferior in most of my images. DxO tended to blow out highlights that NX-D preserved, and DxO visibly muddied very fine detail that came out very clear in the Nikon tool. NX-D also preserved colors better, though those results were inconsistent; DxO sometimes turned a color a little neon.

And I found that Nikon's distortion and chromatic aberration tools were the equal of DxO's. (I didn't use any other of DxO's tools so I can't comment on them.)

There was one exception -- DxO's PRIME noise reduction tool is slow but does an incredible job, and this from someone who's tried many tools including Topaz, Noise Ninja, and others. I've never seen anything better. Unfortunately, there's no way to use PRIME as a standalone tool.

Anyway, that's why I've given up DxO and switched to Nikon's NX-D.

[Kevin Purcell]

DxO Optics Pro has two problems that potential new users should understand.

One general: it doesn't support DNG files from unsupported cameras. So if you don't buy the next version you can't get "new camera support" by repackaging unsupported RAW files into DNGs with Adobe DNG Converter. This trick works with any application that has full DNG support.

One specific: it doesn't support Fujfilm XTRANS sensors. This means there are no DxOmark results for XTRANS cameras.

Other users should check for specific issues e.g. Canon users might be interested in knowing that sRAW and mRAW are not supported.

[Christer]

That really goes for ACR/Lightroom as well. LR4 saw enormous improvement in highlight recovery, which allowed me to fix some bird photos I'd previously deemed beyond repair. Luckily I was too lazy to remove the files from my Lightroom library. These days I just keep the raw files if they have potential as long as they aren't several stops overexposed or severely out of focus. Disk is cheap and software keeps getting better and better.

[dd-b]

I agree that we've entered the fortunate and unfortunate land where different software works better for different images. It's great to be able to improve images from the past; it's less great to have to learn (and pay for) multiple programs to do it.

In my big runthrough of RAW converters a few years ago, DxO was one I tested, and it was the lack of anything approaching decent noise control that made me quickly drop it from consideration. Do you know if they've made big strides in noise control lately? Or do we just have different needs, tastes, or something? (I'm trying to get web-size or very modest print size images from extreme ISOs, 3200 and up.)

[Stephen Scharf]

The only issue I have with DXO is that they don't support Fuji RAF file conversion.

[Henk]

DxO may have it's qualities as raw developer but to advertise DxO as the solution for older raw files is a bridge too far.
I just checked on the DxO website and found that three of my older cameras are not supported. The raw developer I use now does and squeezes a lot more information from the files than the raw developers of 10 years ago.

Merry Christmas
Henk

[Richard Howe]

A handful of my favorite images from my Manhattan Street Corners project of now almost ten years ago were technically flawed in that they were slightly out of focus, or too noisy, or underexposed, or even all three. The best of images from that project could take huge enlargements, up to 90" wide, most of them were good up to 40", but — of course! — among my favorites these were disappointing at any and every size.

I started using PhotoNinja a couple of years ago, and recently got around to seeing what I could do with those "lost" images from ten years ago, and I must say I was amazed. They still wouldn't stand up to being enlarged as much as the good ones, but they're more than adequate at widths up to 24" and at 12" you'd never know there had ever been any problem at all.

I have a few other images from eight to ten years ago that somehow got underexposed by like four stops! Don't ask me how, I really don't know, but with DxO I was able to make something out of them that to my biased eye is even more than "good enough."

(I also got Capture One, but somehow I've never warmed to it, not because of the results, but because I've always found it awkward to use compared to the others.)

So, anyway, yes! Yes! There's a lot of life left in some of the old raw files, and the newer software really is able to draw it out, though sometimes a fair amount of patience and a willingness to try things that "I just know won't work" is required to make it happen.

Best, and Merry Christmas, Mike!

Richard Howe

[John Krill]

I use Lightroom not because the Develop module is so great but the LIBRARY module IS soooo great. Even when I get lazy and forget to insert keywords, etc, I can still find what I want in 9 years and thousands of images of digital photography.

So if you want me to move over to another app then show me what their version of the Library module can do.

And now Merry Christmas to all.

[John Lehet]

In reply to John Krill:

Oh yeah, I use Lightroom too, because the library module is so great. I also love the develop module, but more for the mad-scientist visioning and re-visioning the catalog of images, sorting and sifting, finding the really good ones, sifting again and re-sifting and re-visioning.

For serious stuff it goes out of Lightroom and into Photoshop, after I'm done playing and re-seeing and pushing on things. In Photoshop I apply better (for my purposes) tools like DxO for the raw (at least for my old Nikon and new Olympus files) and Nik for the black and white.

Then I actually go back into Lightroom to print. I have a catalog of files that are tweaked just to print, and I find LR gives me a very good printing workflow.

As for the older cameras mentioned above, sorry for that. I had one Canon point and shoot in early 2004 that is not supported by DxO, but everything I've had after that is supported -- older cameras like some Canon point and shoots, a Nikon D70, D200. The Canon G11 shows maybe the biggest improvement of anything from DxO.

It's true about DxO and the DNG files. I unfortunately had a workflow of converting everything to DNG all along up until I started using DxO. I almost always kept the original NEF and CRW files on CDs and then later DVDs, and now I have all of those on a separate disk which I search when I want to work on something. It's very unfortunate that I did that conversion, and my investment in the Lightroom catalog is all tied to those DNGs up to that point. I did the DNG conversion to save disk space (definitely smaller files) and also for Adobe's promise of a common file utopia. I'm finding the DNGs have yielded none of that advantage, and in fact have turned out to be less good than the original files, something I prepared for but didn't really expect.

The other thing I found while pulling back all the old raw files is that the optical media failed much more often than I would have thought. I used "good" brands, stored them in the dark. I had several discs that would not read on any machine I had. So those NEFs are lost forever as far as I can tell now.

[Moose]

One thing that may be of considerable interest to µ4/3 shooters is the difference in linear distortion correction from what Oly and Panny put in their files.

Both makers relax the requirements for linearity in lens design to allow the lens designers to correct other aspects more fully. They then apply correction in camera for JPEGs and in their Raw converters for Raw files. Adobe has entirely bought into this, exactly, so far as I can tell, replicating the correction of, for example Oly's Viewer 3.

What's odd/interesting about this is that the Oly/Panny/ACR correction works by throwing away the corners/edges. DxO corrects the same images in a quite different way, which results in a considerably wider angle of view. If all that's needed is linear distortion correction, though, PTLens does as well, sometimes better, at a much lower price.

This example shows how much of the original image is lost with Oly/ACR correction.

The next logical question is, which represents the stated focal length?

The other thing I've found is Oly/Panny under correct linear distortion in their standard zooms. Not so obvious in the above example, but very much so here.

You can see that DxO does a much better job, even while retaining the full angle of view. PTLens actually slightly improves on DxO for the Oly 12-50.

-------------------

I'm surprised to see that it's been five years since I compared DxO To ACR/LR and Viewer for color, contrast, noise, fine detail, etc. At that time, I found DxO defaults for color/contrast/curve, whatever, to be inferior to the others.

I notice comparisons of DxO to LR. I've never liked LR for conversion and editing, so my comparisons have been to PS with NeatImage and Focus Magic plug-ins.

[Moose]

Jon Krill says "So if you want me to move over to another app then show me what their version of the Library module can do."

For me, it's simple. LR library to find it, then open in a program I like to work in. \;~)>

The LR Map module is like magic for me. I've geocoded my images for several years. Finding all my images from different times in a particular place using the map is fabulous.

[Kevin Purcell]

@Moose: "What's odd/interesting about this is that the Oly/Panny/ACR correction works by throwing away the corners/edges. DxO corrects the same images in a quite different way, which results in a considerably wider angle of view."

http://www.moosemystic.net/Gallery/tech/DXO/_A001643_dist_corr.htm

Most applications are correcting the barrel distortion by adding inverse pincushion distortion. That will transform the edge of the image from a rectangle to a "pincushioned rectangle" so the edges of the frame will all bow inwards. Then they crop inside that distorted rectangle to get back to a rectangular frame. In the process they throw away the corners and some of the edges.

I suspect that this is done in camera in place by differential stretching the image and placing the output image back in the same sized frame (so you loose the edges). So both uncorrected images and correct images (in camera) have the same dimensions.

This is the correction that Olympus had in mind when designing the camera (perhaps for simplicity in camera) so making the focal length of the lens a bit shorter (17mm or 34mm eq) than the desired value (a bit more than 35mm eq) to compensate for the cropping during distortion correction.

Dxo and PT Lens(?) use a different approach to lens correction. I think they're not trying to deliver an output image with the same dimensions and the input image. It looks like they stretch the image into a different sized box with different shape. You can see the aspect ratio changes as the correction is made. Take a look at Moose's page and you'll see the Dxo corrected image is physically wider (in pixels) and is stretching the edges to make the correction.

Even Dxo is cropping a little and PTLens the least, In fact PTLens is probably not cropping at all. You can see that PTLens leaves a black pincushion shaped border (i.e. there are added black pixels at the middle of the edges). It's more obvious if you magnify the images to inspect them.

After all you can't really apply a lens correction transform to a rectangular image and expect the output to be rectangular. You either accept a non-rectangular image or crop. Or fudge the transformation slightly to be "more rectangular" at the edges. In PTLens' case I think they decided that they'd correct and then let the user decide on the crop afterwards.